Sunday, April 24, 2005

 

GCC 4.0.0 is released

GCC 4.0.0 was released just this past week, what does not mean for the normal computer user. Nothing, really nothing. As people should not be compiling their OS with the newest compiler. Now they can compile their own scientific program with the new compiler to try it out and see if it is faster than the old one.

Every time a release happens and there are still regressions which are not fixed, I know that those bugs will the most duplicated bugs for the first couple weeks. Though the number of new regressions except for i386 SSE/MMX problems due to the rewrite, RTH is handling all of them which is good. Most of the SSE/MMX bugs have been stupid mistakes, wrong constraint on an instruction or we don't want to do a transformation on the RTL level. The only other new regression is tree-ssa-pre copying only part of an instruction which eliminated. Hopefully we can fix these couple of problems and release 4.0.1 within a month (Mark, the RM, said two/three months but I feel we should release often).

Well on 4.1.0 news, well VRP has been finding some latent bugs in fold which is good but looks like VRP and the tcb branch was not as tested as I had hoped for, oh well. Ada does not bootstrap right now because of VRP miscompiling the front-end so it ends up in infinite loop which is bad. There are also some other regressions on the mainline due either to TCB or Jeff's new jump threader. There is a major regression which causes GCC on ppc-darwin not to be able to compile it self and Honza has not fixed it yet, oh well.

Comments: Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?